| UST Faculty Research Colloquium |
Tuesday, October 1, 2019 12:15 PM to 2:00 PM
Improving Long Term Forecasting
Performance of Variance
The variance risk premium (VRP) is defined as the difference between expected return variance under the risk-neutral measure and the expected return variance under the objective probability measure. We formulate a GARCH-type model for the conditional variance of the consumption growth rate variance and show that, in a representative agent model with Epstein Zin-Weil recursive preferences, the conditional variance of VRP (Var-VRP) appears explicitly in the equity risk premium. We test and find that, while the predictive power of the VRP wanes at around the 4-month horizon, the predictive power of Var-VRP remains significant between the 4- to 15-month horizons, complementing and boosting the adjusted-R2 of VRP. Adding well-known traditional fundamental variables further increases the significance of Var-VRP. A back-test using a simple binary strategy predicted by Var-VRP and VRP improves the Sharpe ratio performance by 17.7% over the model using VRP alone, and almost doubles the Sharpe ratio of holding S&P 500 only.